remorseful - Story 116316 | Emotion Vectors in Gemma4-31B

He’d set the table, a simple meal, but the atmosphere felt charged. He was attempting to speak, but the words felt clumsy. He was trying to say something. In her native tongue, Farsi. She closed her eyes, and a single tear traced a path down her cheek.

She had always hidden her true self. Built a fortress of silence. Now she was trapped in it. Seeing him struggle with the language, and then recognizing that he'd been trying to get closer to her family, and to know her, was too much. She’d been such a coward.

Emotion: remorseful

Cluster: Shame / Guilt

PC1 (Valence): -1.57 Negative

PC2 (Disposition): 1.09

Role in Research

This story is one of 1,000 stories generated for the emotion remorseful. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 remorseful stories, after denoising with neutral dialogue baselines, produces the remorseful emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the remorseful vector is projected through the unembedding matrix.

Promoted:

`S`	0.656
`L`	0.580
`悔`	0.463
`remorse`	0.416
`been`	0.398

Suppressed:

`de`	-0.444
`(!)`	-0.309
`имеется`	-0.289
`oldukça`	-0.287
`😍`	-0.285